Optimal congestion bit setting in a flow control scheme using neural networks

We describe a neural network-based technique for optimal congestion bit setting in a binary feedback flow control scheme for computer networks. This technique employs the sensitivity of the system performance to generate feedback from the network to the data sources. The optimal direction for rate adjustment at the source is based on a single bit feedback signal from the network which depends upon the sign of the sensitivity of the system performance index with respect to the network queue input rate. Simulation results are presented to show the performance of this gradient-based technique compared to the conventional queue-based approach for congestion detection